NRLgate -
Plagiarism by Peer Reviewers
Sections 4.3 thru 4.5
This page is part of the NRLgate Web site presenting evidence of
plagiarism among scientific peer reviewers involving 9 different peer review
documents of 4 different journal and conference papers in the fields of
evolutionary computation and machine learning.
This page contains sections 4.3 through 4.5 of "Evidence of plagiarism
in reviews #1, #2, and #3 of a paper on electrical circuit design submitted
to the Evolutionary Computation journal."
Go to top of NRlgate Home Page
Go to Abbreviated Table of Contents
Go to Detailed Table of Contents
Go to Complaint Letter
to the Evolutionary Computation journal
Go to top of Previous Page
Go to top of Next Page
4.3. The unlikely coincidence that reviewers #1 and #2 would both make
the mistaken assertion that the paper lacked a specific item even though
the paper actually devotes almost 2 pages to the item
Most published experimental papers in the field of evolutionary computation
and machine learning contain a "baseline" comparison of the performance
of the paper's technique to a performance of a "blind random search"
of the search space of the problem. The reason is that people legitimately
and naturally wonder whether any proposed new automated technique for machine
learning and artificial intelligence performs better than "blind luck"
or the proverbial "monkey at the typewriter" in solving problems.
Two of the three reviewers of the submitted ECJ paper complained about the
absence of the virtually obligatory "baseline" comparison to random
search in my ECJ paper.
Peer reviewer #1 said,
- > ... Given the goal of finding a Lisp expression
> for the system response, how complex is the space being explored
> by GP? How would random search over the same bounded space
of
> Lisp expressions perform? ...
(Emphasis added)
Peer reviewer #2 said,
- ... Since results are
- obtained so quickly (within 50 generations) it is especially important
- to evaluate the density of acceptable solutions in the search space.
- This usually means comparison with some baseline approach, perhaps
- random search
.
- (Emphasis added)
Reviewer #2 was so concerned about a comparison with the absence of the
obligatory comparison with blind random search that he repeated his concern
later in his review.
- The manuscript could be made acceptable without
many further experiments if revised ... [to] ... [d]iscuss the density of
acceptable solutions, and investigate how well random search performs
on the chosen problems.
(Emphasis added)
However, my ECJ paper did have this very kind of "baseline"
comparison!
Moreover, the comparison could not be easily-overlooked --- it occupied
almost two pages of space in my submitted ECJ paper.
This "baseline" comparison could hardly be missed by a peer reviewer
who was relying on his own reading of the paper to formulate his own opinion
about the contents of the paper.
Indeed, the comparison with random search was covered extensively
in the paper from five different perspectives:
- a comparison with blind random search involving 72,000 individuals (without
automatically defined functions) from the precisely the "bounded space
of Lisp expressions" desired by reviewer #1 and "density of acceptable
solutions in the search space" desired by reviewer #2,
- a comparison with blind random search involving 44,000 individuals (with
ADFs) exactly as desired by reviewer #1 and #2,
- a lengthy discussion, with detailed analysis, of examples of the programs
produced by blind random search,
- two figures that address the question of the baseline value for a random
search for the problem, and
- a clear summarizing statement in the paper saying,
- [such and such] ... is the baseline value for a random search of the
space of possible compositions of terminals and primitive functions for
this problem.
Indeed, the baseline comparison in my submitted ECJ paper was so extensive
that I would not have been surprised if one of the three peer reviewers
had criticized my paper for devoting too much space to the baseline comparison.
Between 1988 and 1995, I have submitted about 100 papers on genetic programming
to various peer-reviewed conferences, journals, and edited collections of
papers. Almost 70 have now been published (or have been accepted for publication).
Each of these 100 submissions were reviewed, on average, by 3 peer reviewers
(sometimes by as many as 14). Thus, I have received approximately 300 peer
reviews of my submitted papers on genetic programming over the years. This
accumulation of peer reviews is a not insubstantial sampling of the way
a broad range of anonymous scientific peer reviewers react and comment on
technical papers in this field.
I know of no case in which any of approximately 300 other peer reviews missed
the presence of this baseline comparison and falsely claimed that I had
omitted something that was manifestly in the paper. In other words, the
writers of all of these other 300 reviews had no difficulty in spotting
the presence of this virtually obligatory item in the numerous other papers
that I submitted to other scientific journals. conferences, and edited collections
of papers.
How could two peer reviewers mistakenly assert that my submitted ECJ paper
lacks a certain clearly-identified item when the paper demonstrably devotes
almost two pages to that item? If, hypothetically, a peer reviewer were
prejudiced against a particular technology or author, I can appreciate that
he might hastily read a submitted paper and then construct a deceptive paper
trail of arguable flaws to conceal the obviously illegitimate basis for
his preordained negative decision about the paper. If his review were cursory,
I can appreciate how one such reviewer might conceivably make the gaffe
of asserting the absence of something that the paper actually contains.
Indeed, the first reviewer committed just such a gaffe. If, hypothetically,
a second peer reviewer were also similarly prejudiced, I can also appreciate
how this second reviewer might construct his own hypocritical paper trail
of negative-sounding verbiage to disguise his bias. However, two prejudiced
reviewers would not make the same egregious gaffe if they were acting independently
in accordance with fundamental principles of the peer review process.
4.4. The unlikely coincidence that both reviewers #1 and #3 would complain
about "advertising"
There are many different things that the writer of a relatively short review
document can raise about a relatively long submitted paper. There are many
different words that can be used to express the same thought.
As previously mentioned, the accumulation of approximately 300 peer reviews
of my submitted papers on genetic programming over the years constitutes
a not insubstantial sampling of the way a broad range of anonymous scientific
peer reviewers react and comment on technical papers that they review in
this field. In almost 100 submitted papers on genetic programming between
1988 and 1995, I have used an almost-identical introductory paragraph to
summarize genetic programming. Like many people today, I use a word processor
to write my own articles and I "cut and paste" this standard "background"
paragraph into each paper (while slightly customizing it). I patterned my
standard paragraph after many similarly constructed paragraphs that I have
seen many times in the literature. Even today, I continue to use a substantially
similar standard paragraph in substantially the same style in virtually
every paper that I submit for publication.
Reviewers #3 and #1 made a non-mainstream, antagonistic comment concerning
forms of the pejorative word "advertising" that I have never seen
(before or since) among the peer reviews of my scientific papers.
Peer reviewer #3 complained,
- The authors clearly state some research goals, but I think
they have
- some "hidden agendas"
that should be made explicit, most
notably
- ...
- to advertise the genetic
- programming metho
d (a full 1 1/2 pages of text is devoted to
- describing other applications that the method has been used for), and
- only slightly more subtly to advertising Koza's book and videotape.
I
- think this last agenda is inappropriate for a scientific journal.
- (Emphasis added).
Peer reviewer #1 complains,
- > The presentation style is very uneven for a
scientific journal article.
- ...
- > Although there are many carefully written sections,
there are parts
- > which:
- ...
- > - are blatant advertisements for the authors' book
and videotape (pp. 4-5).
- >
(Emphasis added).
It is, of course, no offense --- much less a "hidden agenda" ---
for the author of a paper on genetic programming
- to advertise the genetic programming method
Indeed, the primary purpose of most articles in scientific publications
is to "advertise" a successful method or result (using the word
"advertise" in its literal and non-pejorative sense).
How common is it for a scientific peer reviewer to criticize the author
of a paper because he "advertised" his "method"?
We again make reference to the computer file containing the 64,109 words
of the 316 paper review forms from the 86 peer reviewers of the genetic
programming papers at the Genetic Programming 1996 Conference.
The word "advertise" does not appear, in any form, among these
64,109 words written by contemporary peer reviewers in this field.
Of course, there was no "hidden agenda" to
- to advertise the genetic programming method
in the my submitted ECJ paper. That was definitely one item on my "agenda."
as, indeed, it is on the agenda of most authors who write papers..
Also, it is, of course, no offense --- much less a "hidden agenda"
--- for the author of a scientific paper to provide
- a full 1 1/2 pages of text is devoted to describing other applications
that the method has been used for
Of course, references to previous applications of a technique discussed
in a scientific paper are mandatory in scientific papers. A detailed recital
is especially appropriate for a new technique such as genetic programming.
Even today, it is not universally accepted that genetic programming can
solve a wide variety of problems over a wide variety of fields (and it certainly
was not universally accepted at the time of submission of this paper). There
is no "hidden agenda" to "advertise the genetic programming
method" by providing citations in the scientific literature to problems
that genetic programming has successfully solved. Indeed, that was definitely
one item on my "agenda."
As to the videotape, the idea of having a videotape to accompany a book
that presents a series of two dozen segments of animated computer runs was
unique at the time of submission of this paper to this journal in the fields
of genetic algorithms, machine learning, and artificial intelligence. Indeed,
this approach is still unique, to my knowledge, in scientific publishing,
in general. My videotape contains about two dozen segments of actual computer
runs and they provide a "live" visualization of material that
can be difficult to understand in text form. A picture is worth a thousand
words. It would be preposterous for me suppress the existence of the videotape
(or to hide the existence of my own book) in a paper intended to communicate
information to readers of a scientific journal. This tape (and the similar
videotape accompanying my second book) are, I believe, a contribution to
scientific publishing that will become increasing common in future years.
Let's return now to the common features of the reviews #1 and #3.
Both reviewers #1 and #3 expressed a thought not normally seen in the texts
scientific peer reviews; they both articulated their non-mainstream interpretation
using the same specific pejorative word (when many other phrases or words
could have chosen to express the same thought).
Reviewer #1 complains about the reference to both
- the authors' book and videotape
while reviewer #3 complains about the reference to both
- Koza's book and videotape
Reviewer #1 complains that the "presentation" is
- uneven for a scientific journal
while reviewer #3 complains that the "presentation" is
- inappropriate for a scientific journal
Notice the symmetric "on the one hand, on the other hand" construction.
Reviewer #1 said,
- The authors clearly state some research goals,
but I think they have
- some "hidden agendas"
Reviewer #1 said,
- > Although there are many carefully
written sections, there are parts
- > which:
- ...
- > - are blatant advertisements
- (Emphasis added in each case above)
Of course, these particular two reviewers are entitled to their personal
non-mainstream opinion about the matter of what is, or is not, "blatant"
"advertising." The point here is not whether my paper really and
truly is or is not "advertising." The point is that, over a six
year period, these two reviews from the ECJ are the only ones (out of about
300 peer review documents) that ever criticized this standard paragraph
as "advertising" and getting one (much less two) such reviews
on the same paper is a rare and improbable convergence of events in the
world of peer reviews.
4.5. The unlikely coincidence that both reviewers #1 and #2 would abbreviate
the paper's title and that they would both abbreviate it to the same 8 words
Peer reviewer #3 refers to my submitted ECJ paper by its full 15-word title:
- Use of Automatic Function Definition in Genetic Programming to Facilitate
Finding an Impulse Response Function
However, reviewer #1 and reviewer #2 abbreviate the title to the same 8
words.
How common is the practice of abbreviating the author's title on a paper
review form?
We again make reference to the computer file containing the 64,109 words
of the 316 paper review forms from the 86 peer reviewers of the genetic
programming papers at the Genetic Programming 1996 Conference.
There were 27 reviews of papers whose titles contained between 11 and 17
words. In 63% of the cases, the reviewer entered the author's full exact
title onto the paper review form without alteration. Thus, the probability
of randomly drawing 2 reviewers who abbreviate the author's title is about
1 in 9.
More importantly, there are many different ways to abbreviate a title. In
the case of the 10 abbreviations, there was no case where two reviewers
abbreviated a particular title to the same words (or even converged to the
same number of words). In fact, it was somewhat surprising to see the many
different ways for abbreviating titles that were used by contemporary peer
reviewers in this field. For the abbreviated titles, no two styles were
the same. The different styles for shortening titles included
- using just the first 2 words of the author's title
- using the first coherent phrase
- using a coherent phrase from a later part of the title
- dropping articles, such as "the" and "an"
- abbreviating "genetic programming" to "GP"
- abbreviating an especially long word
- adding "etc." in lieu of deleted words
- using ellipsis (3 dots) in lieu of deleted words
- using a long dash in lieu of deleted words
- leaving the "title" line entirely blank on the paper review
form
We do not have sufficient data to compute the joint probability of convergent
abbreviation to identical words; however, a rough guess is that it is around
1 in 100.
Without the suggestive power of one already written review, isn't it improbable
that reviewers X and Y would both abbreviate my title and that they would
also converge to the identical shortened version?
This is not the only occasion when we encounter two peer reviewers of the
same paper both belonging to the small minority of peer reviewers who abbreviate
the author's title and who coincidentally converge on the same abbreviation.
As will been seen elsewhere, reviewers X and Y abbreviated the 14-word title
of my paper on optimal control strategies submitted to the Machine Learning
Conference and they both converged to the same 6 words. See
section 3.8.
Author: John R. Koza
E-Mail: NRLgate@cris.com
Go to top of NRlgate Home Page
Go to Abbreviated Table of Contents
Go to Detailed Table of Contents
Go to Complaint Letter
to the Evolutionary Computation journal
Go to top of Previous Page
Go to top of Next Page